Content of the invention
The purpose of the present invention is to gather palmprint image under multiple spectrum irradiates, and the palmmprint figure by these multispectral collections
The fine lines of multi-light spectrum palm print as obtaining extremely precisely palmmprint lines after processing extracts recognition methods.
Step of the present invention is:
A, multi-optical spectrum image collecting:
For the same area of palm, under each spectrum, gather a width palmprint image;Obtain multispectral image,,
...,, wherein a, b, c ..., n be 1,2,3 ..., n difference spectrum;
The fine patterned feature of b, single-spectral images extracts:
First, to the image under certain spectrumThe contourlet conversion carrying out three layers of direction transformation redundancy is decomposed, wherein, obtain 15 width sub-band images, wherein,Represent certain spectrum pictureAfter decomposing
Lowest frequency sub-band coefficients,WithRepresent " level " direction high-frequency sub-band system that ground floor decomposites respectively
Number and " vertical " direction high-frequency sub-band coefficient,Represent all high-frequency sub-band coefficients that second and third layer decomposites;Its
Secondary, design factor matrixWithAbsolute value matrix, respectively by each yuan in two absolute value matrixs
Element arranges from big to small, and the value taking out middle element is remembered respectivelyWith;Then, travel through coefficient matrix's
Each element, when element value is more than or equal toWhen, illustrate that this is strong edge coefficient, then retain the value of element, when element value is little
InWhen, illustrate that this is weak fringing coefficient or noise, then the value of element is set to 0, eliminated, in the same manner to coefficient matrixProcessed;Finally, lowest frequency sub-band coefficientsThe all high-frequency sub-band decompositing with second and third layer
CoefficientAll set to 0;
The fine patterned feature of c, multispectral image merges:
By the image to be fused under different spectrum、、、…After processing through step b, respective ground floor " level "
Direction high-frequency sub-band coefficient、、…、Blend, respective ground floor " vertical " direction high frequency
Sub-band coefficients、、…、Blend;The fusion of ground floor high frequency coefficient adopts absolute coefficient
The maximum fusion rule of choosing, that is, the big coefficient of absolute value choosing coefficient on correspondence position is as the coefficient after merging;Second and third layer
High frequency coefficient merges and lowest frequency coefficient merges the fusion rule all adopting coefficient superposition, that is, the coefficient chosen on correspondence position adds
As the coefficient after merging, after merging here, their coefficient is 0 to value preset;Each layer coefficients travel direction conversion after merging
The inverse transformation of redundancy contourlet;
D, Morphological scale-space:
First, binaryzation, after direction transformation redundancy contourlet inverse transformation, each pixel value of image is that have just to have
Negative coefficient, the coefficient being wherein more than 1 is shown as white on image, and the coefficient less than 0 is shown as black, coefficient on image
Value appears dimmed on image in the coefficient between 0~1, therefore, all pixels value is less than the value zero setting of 0 pixel, will
The value that all pixels value is more than or equal to zero pixel puts 1, obtains the image after binaryzation;
Then, image negates, with 1 image deducting binaryzation, black picture element in image is converted to white, will scheme
In picture, white pixel is converted to black, that is, obtain the image after inverseLines is white, and other are black;
Finally, morphological dilations, refinement, expansion process, using structural element ' disk ', radius is 1, to imageEnter
Row expands, the image after being expanded;By the image after expandingIt is refined into single pixel image;Due to single pixel imageThe place intersecting in streakline is likely to occur null point, therefore will it be expanded, the structural element of expansion is ' disk ' again, becomes
Footpath is 1, obtains the image after microdilatancy,The pixel wide of middle streakline is about 3 pixels;Wherein expand refer to by
Black region in image becomes big, and white streakline attenuates;
E, cross line and the identification of M shape line:
(1) search for imageIn white pixel point, when the point of white pixel point m1 eight neighborhood is white, then this point
It is the interior point of streakline, as each streakline central intersection point;
(2) the size and shape feature according to cross pattern, sets up " form matrix ", and wherein cross shape shape matrix is type,
The positive cross matrix of 45 ° of states is type;
(3) matrix formed by pixel point value in the region centered on m1 is set to m1, by m1 respectively with different types of
" form matrix " carries out AND operation, and the matrix of consequence obtaining is asked with the value preset of matrix, using this value preset as " m1 point be i ii type
The marking value in cross searching crosspoint ";
(4) when marking value is closer to the value preset of " form matrix ", show point in the rectangular area centered on m1
It is distributed closer to " form matrix ", conversely, then lower with the similarity of " form matrix ";
(5) requirement according to accuracy of identification arranges threshold value t, when marking value is less than or equal to t, abandons this point, continues to search
Rope next one white pixel point carries out " form matrix " contrast;When marking value is more than t, show that this point m1 is central intersection point,
And shape is with " form matrix ", indicate that this point is cross pattern central intersection point;When same pixel, it is " in i type cross
Heart crosspoint ", is " ii type cross searching crosspoint " again, then show that this point is " M shape line central intersection point ".
The fine lines of multi-light spectrum palm print of the present invention extracts the image collecting device of recognition methods: base is by cross bar and montant
Constitute, vertical rod is provided with base;
Sideslip groove is had on cross bar, cross motor is placed in sideslip groove, the axle of cross motor is connected with stringer leading screw, stringer
Stringer support has been threaded connection on leading screw;
Vertical runner is had on montant, longitudinal motor is placed in vertical runner, the axle of longitudinal motor is connected with row leading screw, row
Row support has been threaded connection on leading screw;
Row support is fixedly mounted on stringer support;
Vertical pin is provided with row support, vertically pin is plugged on object disposing platform, and object disposing platform is placed on lower drawbar
On, lower drawbar is fixed in vertical rod;
Lifting bracket is provided with object disposing platform, camera is placed in inside lifting bracket;
Annular cover is connected with lens barrel by screw thread, the annular array that the arrangement of the LED lamp of different wave length uniformly interlocks
Be fixed on annular cover lid medial surface, light source control box is fixed on the lateral surface of annular cover, light source control box built with
The battery powered to LED lamp and the switch controlling light source switching, the multispectral mirror of annular cover, lens barrel and light source control box composition
Cylinder is mounted on lifting bracket, and lens barrel center is aligned with the optical center of camera;
Drawbar is installed with vertical rod, upper drawbar is placed with background board.
The technical scheme that the present invention provides proposes a kind of multi-light spectrum palm print fine lines extracting method, and that is, direction transformation is superfluous
The feature extracting method that remaining contourlet conversion is combined with Morphological scale-space, merges the feature of multiple spectrum pictures, more
Mended the palmmprint line that existing method extracts have the shortcomings that interruption discontinuous it is impossible to extract tiny palmmprint line well, we
Method can not only be distinct display palm on main line, and shallow and thin fine lines can be extracted;Propose a kind of fine lines to know
Other method, that is, pass through to construct " form matrix " method to the point marking on streakline, identify the fine line with special shape
Road cross line and M shape line;The multi-light spectrum palm print fine lines picking platform of the present invention can be according on studied palm
Region different, flexibly move the zones of different on be aligned palm and carry out adopting of palmmprint fine lines image under multiple spectrum
Collection, more apparent than the image that existing multispectral harvester collects, the region of collection is more flexible.
Specific embodiment
The step of the present invention is:
A, multi-optical spectrum image collecting:
For the same area of palm, under each spectrum, gather a width palmprint image;Obtain multispectral image,,
...,, wherein a, b, c ..., n be 1,2,3 ..., n difference spectrum;
The fine patterned feature of b, single-spectral images extracts:
First, to the image under certain spectrumThe contourlet conversion carrying out three layers of direction transformation redundancy is decomposed, wherein, obtain 15 width sub-band images, wherein,Represent certain spectrum pictureThrough decomposing
Lowest frequency sub-band coefficients afterwards,WithRepresent " level " direction high frequency that ground floor decomposites respectively
Band coefficient and " vertical " direction high-frequency sub-band coefficient,Represent all high-frequency sub-band systems that second and third layer decomposites
Number;Secondly, design factor matrixWithAbsolute value matrix, respectively will be each in two absolute value matrixs
Individual element arranges from big to small, and the value taking out middle element is remembered respectivelyWith;Then, travel through coefficient matrixEach element, when element value is more than or equal toWhen, illustrate that this is strong edge coefficient, then retain the value of element,
When element value is less thanWhen, illustrate that this is weak fringing coefficient or noise, then the value of element is set to 0, eliminated, right in the same manner
Coefficient matrixProcessed;Finally, lowest frequency sub-band coefficientsAll with what second and third layer decomposited
High-frequency sub-band coefficientAll set to 0;
The fine patterned feature of c, multispectral image merges:
By the image to be fused under different spectrum、、、…After processing through step b, respective ground floor " level "
Direction high-frequency sub-band coefficient、、…、Blend, respective ground floor " vertical " direction high frequency
Sub-band coefficients、、…、Blend;The fusion of ground floor high frequency coefficient adopts absolute coefficient
The maximum fusion rule of choosing, that is, the big coefficient of absolute value choosing coefficient on correspondence position is as the coefficient after merging;Second and third layer
High frequency coefficient merges and lowest frequency coefficient merges the fusion rule all adopting coefficient superposition, that is, the coefficient chosen on correspondence position adds
As the coefficient after merging, after merging here, their coefficient is 0 to value preset;Each layer coefficients travel direction conversion after merging
The inverse transformation of redundancy contourlet;
D, Morphological scale-space:
First, binaryzation, after direction transformation redundancy contourlet inverse transformation, each pixel value of image is that have just to have
Negative coefficient, the coefficient being wherein more than 1 is shown as white on image, and the coefficient less than 0 is shown as black, coefficient on image
Value appears dimmed on image in the coefficient between 0~1, therefore, all pixels value is less than the value zero setting of 0 pixel, will
The value that all pixels value is more than or equal to zero pixel puts 1, obtains the image after binaryzation;
Then, image negates, with 1 image deducting binaryzation, black picture element in image is converted to white, will scheme
In picture, white pixel is converted to black, that is, obtain the image after inverseLines is white, and other are black;
Finally, morphological dilations, refinement, expansion process, using structural element ' disk ', radius is 1, to imageCarry out
Expand, the image after being expanded;By the image after expandingIt is refined into single pixel image;Due to single pixel image
The place intersecting in streakline is likely to occur null point, therefore will it be expanded again, the structural element of expansion is ' disk ', reducing
For 1, obtain the image after microdilatancy,The pixel wide of middle streakline is about 3 pixels;Wherein expand and refer to scheme
Black region in picture becomes big, and white streakline attenuates;
E, cross line and the identification of M shape line:
(1) search for imageIn white pixel point, when the point of white pixel point m1 eight neighborhood is white, then this point
It is the interior point of streakline, as each streakline central intersection point;
(2) the size and shape feature according to cross pattern, sets up " form matrix ", and wherein cross shape shape matrix is type,
The positive cross matrix of 45 ° of states is type;
(3) matrix formed by pixel point value in the region centered on m1 is set to m1, by m1 respectively with different types of
" form matrix " carries out AND operation, and the matrix of consequence obtaining is asked with the value preset of matrix, using this value preset as " m1 point be i ii type
The marking value in cross searching crosspoint ";
(4) when marking value is closer to the value preset of " form matrix ", show point in the rectangular area centered on m1
It is distributed closer to " form matrix ", conversely, then lower with the similarity of " form matrix ";
(5) requirement according to accuracy of identification arranges threshold value t, when marking value is less than or equal to t, abandons this point, continues to search
Rope next one white pixel point carries out " form matrix " contrast;When marking value is more than t, show that this point m1 is central intersection point,
And shape is with " form matrix ", indicate that this point is cross pattern central intersection point;When same pixel, it is " in i type cross
Heart crosspoint ", is " ii type cross searching crosspoint " again, then show that this point is " M shape line central intersection point ".
The fine lines of multi-light spectrum palm print of the present invention extracts the image collecting device of recognition methods: base is by cross bar 3 and perpendicular
Bar 21 is constituted, and is provided with vertical rod 1 on base;
Sideslip groove 5 is had on cross bar 3, cross motor 7 is placed in sideslip groove 5, the axle of cross motor 7 is connected with stringer leading screw
6, stringer leading screw 6 has been threaded connection stringer support 8;Cross motor 7 can only carry out transverse shifting along sideslip groove 5;
Vertical runner 11 is had on montant 21, longitudinal motor 12 is placed in vertical runner 11, and the axle of longitudinal motor 12 is connected with horizontal stroke
Row leading screw 9, row leading screw 9 has been threaded connection row support 10;Longitudinal motor 12 can only carry out longitudinal direction along vertical runner 11
Slide;
Row support 10 is fixedly mounted on stringer support 8;
Vertical pin 13 is provided with row support 10, vertical pin 13 is plugged on object disposing platform 2, and object disposing platform 2 is placed
On lower drawbar 14, lower drawbar 14 is fixed in vertical rod 1;
Lifting bracket 16 is provided with object disposing platform 2, camera 15 is placed in inside lifting bracket 16;
Annular cover 17 is connected with lens barrel 19 by screw thread, the annular that the arrangement of the LED lamp of different wave length uniformly interlocks
Array is fixed on annular cover 17 lid medial surface, and light source control box 4 is fixed on the lateral surface of annular cover 17, light source control
Box 4 is built with the battery powered to LED lamp and the switch controlling light source switching, annular cover 17, lens barrel 19 and light source control box 4
The multispectral lens barrel of composition is mounted on lifting bracket 16, and lens barrel 19 center is aligned with the optical center of camera 15;
Drawbar 18 is installed with vertical rod 1, upper drawbar 18 is placed with background board 20.
Below in conjunction with accompanying drawing, the present invention is done with further detailed description: (following instance is to enter using six width spectrum pictures
Row explanation)
A, multi-optical spectrum image collecting:
For certain same area of palm, under each spectrum, gather a width palmprint image;Intercept this multiple image respectively
Same rectangular area, obtain multispectral image,,...,(a, b, c ..., n be spectrum number);
The fine patterned feature of b, single-spectral images extracts:
First, to the image under certain spectrum() carry out 3 layers of direction transformation redundancy
Contourlet conversion (nsct conversion) decomposes, obtain (- 1) width sub-band images, wherein,Represent certain spectrogram
PictureLowest frequency sub-band coefficients after decomposing,WithRepresent the 1st layer of " level " side decompositing respectively
To high-frequency sub-band coefficient and " vertical " direction high-frequency sub-band coefficient,Represent the 2nd, 3 layers of all high frequency decompositing
Band coefficient;Secondly, design factor matrixWithAbsolute value matrix, respectively by two absolute value matrixs
In each element arrange from big to small, take out in the middle of the value of element remember respectivelyWith;Then, travel through coefficient matrixEach element, when element value is more than or equal toWhen, illustrate that this is strong edge coefficient, then retain the value of element,
When element value is less thanWhen, illustrate that this is weak fringing coefficient or noise, then the value of element is set to 0, eliminated, right in the same manner
Coefficient matrixProcessed;Finally, lowest frequency sub-band coefficientsAll high with what the 2nd, 3 layers decomposited
Frequency sub-band coefficientsAll set to 0.
The fine patterned feature of c, multispectral image merges:
By the image to be fused under different spectrum、、、…After processing through step b, respective 1st layer " level "
Direction high-frequency sub-band coefficient、、…、Blend, respective 1st layer of " vertical " direction high frequency
Sub-band coefficients、、…、Blend;1st layer of high frequency coefficient fusion adopts absolute coefficient
The maximum fusion rule of choosing, that is, the big coefficient of absolute value choosing coefficient on correspondence position is as the coefficient after merging;2nd, 3 floor heights
Frequency coefficient merge and lowest frequency coefficient merge all adopt coefficient superposition fusion rule, that is, choose correspondence position on coefficient add with
As the coefficient after merging, after merging here, their coefficient is 0 to value.
Each layer coefficients after merging are carried out nsct inverse transformation.
D, Morphological scale-space:
First, binaryzation.After nsct inverse transformation, each pixel value of image is that have just to have negative coefficient, is wherein more than 1
Coefficient white is shown as on image, the coefficient less than 0 is shown as black on image, coefficient between 0~1 for the coefficient value
Image appears dimmed, therefore, all pixels value is less than the value zero setting of 0 pixel, all pixels value is more than or equal to
The value of zero pixel puts 1, obtains the image after binaryzation.
Then, image negates.With 1 image deducting binaryzation, black picture element in image (pixel value is 0) is converted into
White (pixel value is 1), white pixel in image (pixel value is 1) is converted to black (pixel value is 0), that is, obtains anti-
Image after colorLines is white, and other are black.
Finally, morphological dilations, refinement, expansion process.Using structural element ' disk ', radius is 1, to imageCarry out
Expand, the image after being expanded;By the image after expandingIt is refined into single pixel image;Due to single pixel image
The place intersecting in streakline is likely to occur null point, therefore will it be expanded again, the structural element of expansion is ' disk ', reducing
For 1, obtain the image after microdilatancy,The pixel wide of middle streakline is about 3 pixels.
E, cross line and the identification of M shape line:
(1) search for imageArea-of-interest in white pixel point, when the point of certain white pixel point m1 eight neighborhood is equal
During for white, then this point is the interior point of streakline, is also the central intersection point of doubtful cross pattern or rice character design.
(2) the size and shape feature according to cross pattern and rice character design, sets up " form matrix " of 11*11.Here set up
" i type cross shape matrix " and " type cross shape matrix ".
(3) centered on m1, take the rectangular neighborhood of 11*11, matrix formed by the pixel point value in this region is set to m1, will
M1 carries out AND operation with different types of " form matrix " respectively, and the matrix of consequence obtaining is asked with the value preset of matrix, by this and
Value is as the marking value of " m1 point be i ii type cross searching crosspoint ".
(4) when marking value is closer to the value preset of " form matrix ", show point in the rectangular area centered on m1
It is distributed closer to " form matrix ", conversely, then lower with the similarity of " form matrix ".Therefore, the requirement according to accuracy of identification
Certain threshold value t is set, when marking value is more than t, shows that this point m1 is central intersection point, and shape is with " form matrix ";Work as marking
When value is less than or equal to t, the next black pixel point of search, execution step (1)~(4) again;When same pixel, it is
" i type cross searching crosspoint ", is " ii type cross searching crosspoint " again, then show that this point is " M shape line central crossbar
Point ".
The fine lines of a kind of multi-light spectrum palm print that the present invention provides extracts the flow chart of recognition methods as shown in Figure 1.
During concrete implementation, for step a multi-optical spectrum image collecting:
The five fingers open and are placed in above multispectral lens barrel user's palm of the hand naturally down, and the back of the hand againsts the downside of background board, in
Refer to refer at root, block gim peg a with the third finger, against gim peg b at the wrist of little finger of toe side, the height of adjustment background board.Keep
The attitude of user's hand is constant, controls two spindle motor coordinated movements of various economic factors so that the digital camera visual field is aligned interested on user's palm
Region, examines determination user's palmmprint area-of-interest by personal computer, is cut using the light source switching switch in light source control box
Change different spectrum, gather and store the palmmprint multispectral image of unified area-of-interest under different spectrum.
Accompanying drawing 2 is the orange light that this multi-light spectrum palm print picking platform can collect, green glow, purple light, ruddiness, and blue light, under white spectrum
Palmprint image.Accompanying drawing 3 is to intercept identical rectangular area in accompanying drawing 2 multispectral image.Accompanying drawing 4 is the user of picking platform collection
The multispectral image of palm zones of different.
For step b single-spectral images, fine patterned feature extracts:
It is described as follows in conjunction with accompanying drawing 5.
(1) to the image under orange light and purple light spectrumWithAll carry out the contourlet of 3 layers of direction transformation redundancy respectively
Conversion (one kind of nsct conversion) is decomposed, and obtains 15 width sub-band images, wherein,Represent orange light imageAfter decomposing
Lowest frequency sub-band coefficients,WithRepresent the 1st layer of " level " direction high-frequency sub-band system decompositing respectively
Number and " vertical " direction high-frequency sub-band coefficient,Represent the 2nd, 3 layers of all high-frequency sub-band coefficient decompositing;In the same manner
Obtain purple spectrum picture's、、With.
Why carry out 3 layers of decomposition, be because if using 2 layers of decomposition, decomposing in the 1st layer of sub-band images obtaining and reflect
Lines meticulous, being unfavorable for the fusion recognition in later stage, if decomposed using 4 layers or more, decomposing the 1st layer of sub-band images obtaining
The lines of middle reflection is excessively thick, loses the information of too many fine line and fold, just extracts the faint fine lines of palm of not selling
, therefore final choice use 3 layers of direction transformation redundancy contourlet conversion decompose, for faint fine line on palm
The effect of the extraction identification on road is best.
(2) calculate orange light image" level " direction high-frequency sub-band coefficient matrix that 1st layer decompositesWith
" vertical " direction high-frequency sub-bandAbsolute value matrix, respectively by each element in two absolute value matrixs from big to small
Arrangement, the value taking out middle element is remembered respectivelyWith;Then, travel through coefficient matrixEach element,
When element value is more than or equal toWhen, illustrate that this is strong edge coefficient, then retain the value of element, when element value is less thanWhen,
Illustrate that this is weak fringing coefficient or noise, then the value of element is set to 0, eliminated;In the same manner to " level " direction high-frequency sub-band
Coefficient matrixProcessed;Finally, lowest frequency sub-band coefficientsAll high with what the 2nd, 3 layers decomposited
Frequency sub-band coefficientsAll set to 0.
(3) calculate purple light image" level " direction high-frequency sub-band coefficient matrix that 1st layer decompositesWith
" vertical " direction high-frequency sub-bandAbsolute value matrix, respectively by each element in two absolute value matrixs from big to small
Arrangement, the value taking out middle element is remembered respectivelyWith;Then, travel through coefficient matrixEach element,
When element value is more than or equal toWhen, illustrate that this is strong edge coefficient, then retain the value of element, when element value is less thanWhen,
Illustrate that this is weak fringing coefficient or noise, then the value of element is set to 0, eliminated;In the same manner to " vertical " direction high-frequency sub-band
Coefficient matrixProcessed;Finally, lowest frequency sub-band coefficientsAll high with what the 2nd, 3 layers decomposited
Frequency sub-band coefficientsAll set to 0.
The experimental result of step b as shown in Figure 5, represents experiment orange light image, experiment purple light image, orange light successively
The 1st layer of " level " direction high-frequency sub-band figure of image, the 1st layer of " level " direction high-frequency sub-band figure of purple light image, the 1st layer of orange light image
" vertical " direction high-frequency sub-band figure, the 1st layer of " vertical " direction high-frequency sub-band figure of purple light image.
For step c multispectral image, fine patterned feature merges specifically:
It is described as follows in conjunction with accompanying drawing 6.
Represent the corresponding conversion coefficient of image after fusion treatment.
During (1) the 1st layer of fusion
Maximum fusion rule is selected using absolute coefficient, that is, chooses the big coefficient conduct of the absolute value of coefficient on correspondence position
Coefficient after fusion, formula as shown in formula 1-1 and 1-2,WithRepresent the 1st layer of " water after merging respectively
Flat " direction high-frequency sub-band coefficient and " vertical " direction high-frequency sub-band coefficient.
(formula 1-1)
(formula 1-2)
(2) when lowest frequency subband and the 2nd, 3 layers of high-frequency sub-band merge
Due to setting to 0 the coefficient of these subbands in stepb, only need to simply be added when therefore merging here
Can.
(3) nsct inverse transformation
By the coefficient after fusion and processCarry out nsct inverse transformation.
The experimental result of step c such as accompanying drawing 6.
Step D-shaped state is processed:
First, binaryzation.After nsct inverse transformation, each pixel value of image is that have just to have negative coefficient, is wherein more than 1
Coefficient white is shown as on image, the coefficient less than 0 is shown as black on image, coefficient between 0~1 for the coefficient value
Image appears dimmed, therefore, all pixels value is less than the value zero setting of 0 pixel, all pixels value is more than or equal to
The value of zero pixel puts 1, obtains the image after binaryzation.
Then, image negates.With 1 image deducting binaryzation, black picture element in image (pixel value is 0) is converted into
White (pixel value is 1), white pixel in image (pixel value is 1) is converted to black (pixel value is 0), that is, obtains anti-
Image after colorLines is white, and other are black.
Finally, morphological dilations, refinement, expansion process.Using structural element ' disk ', radius is 1, to imageEnter
Row expands, the image after being expanded;By the image after expandingIt is refined into single pixel image;Due to single pixel imageThe place intersecting in streakline is likely to occur null point, therefore will it be expanded, the structural element of expansion is ' disk ' again,
Reducing is 1, obtains the image after microdilatancy,The pixel wide of middle streakline is about 3 pixels.
The experimental result of step e such as accompanying drawing 7, represents that binaryzation, image negate, expand, refining, reflation experiment is tied successively
Really.
Existing in order to show that the palmmprint line that multi-light spectrum palm print proposed by the present invention fine lines extracting method is extracted is better than
Method, has carried out contrast test.As shown in Figure 12, (a) is to slap under the natural light that request for utilization number 201310137558 gathers
Print image;B () uses the palmprint image under the orange light and purple light of the collection of multi-light spectrum palm print of the present invention fine lines picking platform;(c)
Extract the palmmprint line of (a) using the most frequently used edge detection operator " sobel " Operator Method;(d) request for utilization number
The method of 201310137558 offers extracts the palmmprint line of (a);E method that () present invention provides is carried out to multispectral image (b)
Lines extracts.Contrast is this it appears that the streakline of method of the present invention extraction is more rich, distinct, clearly links up, discontinuous point is few.
The fine lines identification for step f:
Cross pattern and rice character design identification process figure are as shown in Figure 13.
(1) search for imageArea-of-interest in white pixel point, when the point of certain white pixel point m1 eight neighborhood is equal
During for white, then this point is the interior point of streakline, is also the central intersection point of doubtful cross pattern or rice character design.
(2) the size and shape feature according to cross pattern and rice character design, sets up " form matrix " of 11*11.Here set up
" i type cross shape matrix " and " type cross shape matrix ".
(3) centered on m1, take the rectangular neighborhood of 11*11, matrix formed by the pixel point value in this region is set to m1, will
M1 carries out AND operation with different types of " form matrix " respectively, and the matrix of consequence obtaining is asked with the value preset of matrix, by this and
Value is as the marking value of " m1 point be i ii type cross searching crosspoint ".
(4) when marking value is closer to the value preset of " form matrix ", show point in the rectangular area centered on m1
It is distributed closer to " form matrix ", conversely, then lower with the similarity of " form matrix ";
(5) requirement according to accuracy of identification arranges threshold value t, when marking value is less than or equal to t, abandons this point, continues to search
Rope next one white pixel point carries out " form matrix " contrast;When marking value is more than t, show that this point m1 is central intersection point,
And shape is with " form matrix ", indicate that this point is cross pattern central intersection point;When same pixel, it is " in i type cross
Heart crosspoint ", is " ii type cross searching crosspoint " again, then show that this point is " M shape line central intersection point ".
Accompanying drawing 10 form matrix schematic diagram.Left figure is i type positive cross pattern form matrix schematic diagram, and right figure is 45 degree ten of ii type
Word line form matrix schematic diagram.
The experimental result of step f such as accompanying drawing 8.A part represents " i type cross pattern " center ", the b part expression " ii identifying
Type cross pattern " center ".
The structure of embodiments of the invention multi-light spectrum palm print fine lines picking platform is as shown in accompanying drawing 14 and accompanying drawing 15.
The present invention, by the in the hole of vertical pin insertion object disposing platform correspondence position, digital camera is fixed in object disposing platform
Between, lifting bracket is placed on object disposing platform, and just can frame the position of digital camera.Annular cover passes through screw thread and lens barrel
It is connected, the annular array that the LED lamp arrangement of six groups of different wave lengths uniformly interlocks is fixed on annular cover lid medial surface.Light
Source control box is fixed on the lateral surface of annular cover, and light source control box built with the battery powered to LED lamp and controls light source
The switch of switching.The multispectral lens barrel of annular cover, lens barrel and light source control box composition is mounted on lifting bracket, in its lens barrel
The heart is aligned with the optical center of digital camera, background board take put on support stand and parallel with base.Digital camera collects
Palm print image information carry out the data transfer of real-time synchronization by data wire and personal computer, and seen by personal computer
Examine palmprint image, optionally gather and store palm print image information.Background board is fixed with two towards in the one side of multispectral lens barrel
Individual gim peg, the distribution such as accompanying drawing 11 of two gim pegs.
Can horizontally slip in groove in the middle of one spindle motor internal edge in one side in support stand bottom,
Another spindle motor can horizontally slip in the internal edge intermediate groove of its adjacent side.Two of two spindle motors
Fix with the mesophase of lower carriage and upper bracket respectively on feed screw nut, wherein the bottom being positioned parallel to support stand of lower carriage
The position of face and the leading screw perpendicular to spindle motor, wherein upper bracket also parallel with support stand bottom surface and perpendicular to leading screw electricity
The leading screw of machine.Two single-chip microcomputers are connected with the control line of spindle motor, control the cooperation operating of two spindle motors.
During collection palm print image information, palm is opened by user naturally, and the back of the hand is attached to the downside of background board by the palm of the hand down
Face.Spindle motor drives lower carriage and upper bracket to move together with the vertical pin being fixed on support.Vertically pin is inserted into glove
In the corresponding hole of platform, when spindle motor rotates, drive object disposing platform motion, also just driven the number on object disposing platform
Camera moves.Digital camera lens are fixed in the middle of object disposing platform upward, and lifting bracket is placed on and just frames on object disposing platform
The position of digital camera, multispectral lens barrel is mounted on lifting bracket, and its lens barrel center is aligned with digital camera lens center, the back of the body
Scape plate is placed on support stand and parallel with the base of support stand, and two single-chip microcomputers are connected with two spindle motors control respectively
The cooperation operating of spindle motor processed, personal computer is connected with digital camera, receives the palmprint image letter that digital camera collects
Cease and preserve.The palm of user is motionless, and Single-chip Controlling spindle motor rotates the digital camera fortune driving on object disposing platform
Dynamic, therefore can collect the multi-light spectrum palm print image of diverse location.
The motion control portion of spindle motor:
The motion control schematic diagram of accompanying drawing 9 spindle motor.
The collection lenses of multi-light spectrum palm print picking platform determine specific position by two spindle motors, this two leading screws
Motor is all stepper motor, and control principle is exactly that any point in a plane can be represented by coordinate (x, y).Wherein, x-axis
Correspond to the leading screw of two spindle motors of picking platform bottom with y-axis respectively, this two control leading screws (are assumed by driving stepper motor
Stepper motor a controls x-axis direction, and stepper motor b controls y-axis direction).First, need before setting up model to two stepping electricity
Machine is initialized, and the collection camera lens of multispectral picking platform is partially disposed at (0,0) coordinate;Starting motor a makes it positive
Operating (setting the rotation direction along x-axis positive direction as Positive work) arrive coordinate (0, n), write down step number β of motor operation;Make electricity
Machine a is static, starts motor b Positive work (equally setting the square rotation direction of y-axis as Positive work) and arrives coordinate (m, n), this
When write down motor operation step number be α.If so we make camera if (x, y) place collection image, only step need to be controlled
Stepper motor a is positive to run x* β/n step, makes stepper motor b positive operation y* α/m step, thus can make multispectral picking platform
Any station acquisition image in collection camera lens part moving area.